American Journal of Geographic Information System 2013, 2(1): 1-5

DOI: 10.5923/j.ajgis.20130201.01

Application of Remote Sensing and GIS for Flood Hazard

Management: A Case Study from Sindh Province,
Pakistan
Kabir Uddin* , Deo Raj Gurung, Amarnath Giriraj, Basanta Shrestha

M ountain Environment and Natural Resources Information System, International Centre for Integrated M ountain Development, GPO Box
3226, Kathmandu, Nepal

Abstract Floods are one of the most common hazards in the world, affect ing people’s lives and livelihoods. Flood
hazard mapping and flood shelters suitability analysis are v ital elements in appropriate land use planning for flood-prone
areas. This paper describes application of Remote Sensing (RS) and Geographical Info rmation Systems (GIS) in identifying
flood hazard zones and flood shelters and are therefore important tools for planners and decision makers. The purpose of
this article is to describe a simp le and efficient methodology to accurately delineate flood inundated areas, flood-hazard
areas, and suitable areas for flood shelter to min imize flood impacts. Possible extent of flooding and suitable location flood
shelter sites were modeled and mapped for Sindh Province in Pakistan, using the software ArcGIS model builder. The
output was validated using inundation maps based on flood events that took place in 2010 in Pakistan. These were mapped
using object-based image analysis (OBIA) imp lemented in eCognition software. The catastrophic flood of 2010 inundated a
total area of 7579 km2, while the modeled result indicated the hazard area to be 6216 km2 out of 46138 km2.
Discrepancies in modeled and mapped results are insignificant and acceptable considering the manual flood management
interventions which are beyond the capability of models to represent. Thus, this method is robust enough to develop flood
hazard zoning maps and map shelter sites for flood management.
Keywords Remote Sensing, GIS, OBIA, Flood, Hazard, Sindh River, Pakistan

economic damage worth over US$11 million annually

1. Introduction
More than half the wo rld's populat ion lives in Asia,
which is appro ximately one-fifth of the earth's land area. As
natural disasters increase in both intensity and severity
around the world, the Asian region continues to suffer a
disproportionate number of hazard events and related losses
in lives, infrastructure, stability, and economic progress. In
the period 1994–2004 alone, Asia accounted for one third of
1,562 flood disasters worldwide and nearly 60,000people
were killed in floods[1]. A flood is an overflo w of an
expanse of water that submerges the land. Flood hazards are
the most common and destructive of all natural disasters
and are a constant threat to life and property. Each year,
flood d isasters result in t remendous losses and social
disruption worldwide. Over the past 30 years, floods have
been the most catastrophic natural disaster affect ing, on
average, about 80 million people per year or half of the total
po pu lat io n affect ed by any n atu ral d is aster, caus ing
* Corresponding author:
kuddin@icimod.org (Kabir Uddin)
Published online at http://journal.sapub.org/ajgis
Copyright © 2013 Scientific & Academic Publishing. All Rights Reserved
around the world[2]. As increasing human activity
downstream of rivers results in greater flood damage; floods
themselves are also increasing in size and frequency due to
human activities in the upstream section of the river
system[3]
In recent years, risk-based approaches have received
increasing attention as a viable means to manage flood
hazards. Hazard risk and vulnerability assessment (HRVA)
examines the hazards that may affect a community in order
to determine the risk that each hazard event poses to both
the community as a whole and to vulnerable elements in the
community.
In Pakistan, floods have been recognized as a major
natural calamity. The country has a long history of flooding
fro m the Indus River and its tributaries and the floods of
1928, 1929, 1955, 1957, 1959, 1973, 1976, 1988, 1992,
1995, 1996 and 1997 attest to its destructive nature and its
adverse impact on lives and property in Pakistan. More than
three million homes were destroyed and 160 persons lost
their lives in the great flood of 1973. Three years later in
1976, floods demolished over 10 million houses while 425
lives were lost with other losses amounting to a Rs. 6 billion.
Towards the end of September 1988, an unprecedented
2 Kabir Uddin et al.: Application of Remote Sensing and GIS for Flood Hazard
M anagement: A Case Study from Sindh Province, Pakistan
flood occurred, inflict ing damage nationwide in Pakistan
worth about Rs. 17 billion. The catastrophic flood of 1992
surpassed all previous records with damages estimated at Rs.
50 billion.
Sindh is the south eastern province of Pakistan. The
province is associated with the longest river of Pakistan
which stretches to about 1976 miles (3180 km) long.
Originating in the Tibetan Plateau in the vicinity of Lake
Mansarovar in the Tibet Autonomous Region, the river runs
a course through the Ladakh d istrict of Jammu and Kashmir,
then enters the Northern Areas (Gilg it-Balt istan) flowing
through the North in a southerly direction along the entire
length of the country to merge into the Arabian Sea near the
port city of Karach i in Sindh. It is the major source of
irrigation for Pakistan’s fertile agricultural lands. The
river’s estimated annual flow stands at around 207 cubic
kilo meters (cu km), making it the 21st largest river in the
world in terms of annual flow. Climate is not uniform over
the Indus Basin. It varies fro m subtropical arid and
semi-arid to temperate sub-humid in the plains of the Sindh
and Punjab provinces, and alpine in the mountainous
highlands to the north. Annual precipitation ranges between
100 mm and 500 mm in the lowlands to a maximu m of
2000 mm (water equivalent) on mountain slopes. Snowfall
at higher altitudes (above 2500 m) accounts for most of the
river runoff. Alluvial land in Pakistan is known to be a
flood-prone area.
Earth observation techniques can contribute to finding
out more accurately what causes floods. Together with
flood hazard mapping, earth observation techniques can be
used to assessing damage to property, infrastructure, and
agricultural crops. The uses of remotely sensed data in
identifying the trends of river channel migration and as
source of input data for determin ing river behaviour has
been gaining popularity in recent years. One of the main
characteristics of remote sensing is its capability to generate
a large amount of informat ion frequently and spatially,
Figure 1. Location Map of Study Area
becoming a powerful tool for mon itoring changing aquatic
environments. Remote sensing data has been used to
document water quality estimate water depths and monitor
river channel changes and aquatic habitat[4-6].
Object-based analysis of mu ltispectral imagery was
introduced early on in remote sensing literature[7]; however,
the object-based approach has largely been ignored in
favour of p ixel-based methods which have been easier to
implement[8]. One of the advantages of object-based image
analysis is the multitude of additional in formation that can
be derived fro m image objects co mpared to the amount of
informat ion availab le fro m individual p ixels[9].
This study aims to determine flood hazard, flood extended
areas, and proposed flood shelters using remote sensing and
geographic informat ion systems tools. An effort was made
to prepare a flood hazard zone map and a proposed flood
shelter map using GIS and remote sensing techniques and
incorporating a satellite image, topographic map. Th is study
also attempts to explo it the capabilities of remote sensing
and GIS techniques to suggest an appropriate methodology
for flood hazard mapping.
2. Methodology
2.1. Study Area
The study area is located in the south-eastern part of Nepal
and seven districts of Sindh province (Figure 1). Sindh lies in
a tropical to subtropical region; it is hot in the summer and
mild to warm in winter. Temperatures frequently rise above
46°C between May and August, and drops to a minimu m
average of 2°C during December and January in the northern
and elevated regions. Annual rainfall averages about seven
inches, falling main ly during July and August. The province
is mostly arid, with scant vegetation except for the irrigated
Indus Valley.
 American Journal of Geographic Information System 2013, 2(1): 1-5 3

2.2. Data Used where in formation of single pixels is aggregated into

For this study moderate resolution imaging
spectroradiometer (M ODIS) imagery was used to generate a
flood inundation map. A topographic map and Shuttle Radar
Topography Mission (SRTM ) DEM data fro m various
sources were used to analyse flood hazard area and flood
shelter areas.
2.3. Image analysis
Flood hazard depends on many factors such as flood
inundation and topography of the area and etc. Enu merate
the flood extended area mapping were followed few steps.
The acquired MODIS repro jected UTM, Zone 47. After that
eCognition Developer software was used for object-based
image analysis (OBIA). Figure 2 presents the flow diagram
of object-based image analysis for flood extended mapping.
Co mpared with pixel-based methods, this approach shows
better classification results with higher accuracy as it uses
both spectral and spatial information[4,10,11]. The basic
step in eCognition image analysis is to segmentation a scene
homogeneous image objects. Multiresolution segmentation,
which was used in this study for object-based image analysis,
groups areas of similar pixel values into objects;
homogeneous areas result in larger objects, heterogeneous
areas in s maller ones. Therefore several segmentations were
tested with different parameters until the result was
satisfying.. For each object information on average spectral
values, average Normalized Difference Vegetation Index
(NDVI), and land and water mask were derived. In the
pre-processing stage, an NDVI image was created using the
customised feature in the form: NDVI = (RED – IR) / (RED
+ IR ), NDVI indices values range fro m -1.0 to 1.0,
non-vegetation area negative and vegetation area will have
positive value. Land and Water Mask created by = IR) /
(Green) * 100. Land and water mask index values can range
fro m 0 to 255, but water values typically range 0 to 50. This
informat ion was used to develop suitable classification
algorith ms for flood inundation mapping.

Figure 2. Flow diagram of flood extended mapping

Figure 3. Flow diagram of flood shelter analysis

4 Kabir Uddin et al.: Application of Remote Sensing and GIS for Flood Hazard
M anagement: A Case Study from Sindh Province, Pakistan

The next step was to define classes and insert class Results were selected and the optimu m site identified for the
hierarchy. Image objects were linked to class objects and new flood shelter fro m alternative sites.
each classification lin k was stored in the membership value
of the image object to the linked class. Image objects were
classified using user-defined rules. Objects with an area
smaller than the defined min imu m mapping unit were
merged with other objects. The classified inundation map
was exported to shape file format for further processing and
flood hazard processing and flood shelters and modeling.
Several types of factor/parameter were required to do
flood hazard assessment. Acquired datasets of stream order,
river, dig ital elevation model and land cover including all of
the data that a flood under such circumstances would include.
One of these is slope which is an essential topographic factor
that guides the flood potential of spatial areas. Slope data
layers were generated using the digital elevation model of Figure 4. Flood hazard zoning map
the study area. Vegetation poses a major restraint to flooding;
vegetation reduces runoff and helps in percolation.
Euclidea n allocation calcu lates for each cell the nearest 3. Results
source based on Euclidean distance and calculates the
direction in degrees to the nearest source. Normally, all
nearest source areas are flood-prone areas. Euclidean
allocations of stream order were generated. Euclidean
distances of rivers were generated and short disease from
river ran ked flood potential. In order to co mbine the datasets,
they must be set to a common scale. Each dataset data is then
reclassified to rank the potential hazard location. Then
reclassify date sets came to a common scale ranges from 1 to
5. Each raster was assigned ranks at the time of Legend
reclassification: a ranking of ‘1’ means with the most flood Restricted
potential, and ‘5’, least flood potential. Once, all the four Not suitable
raster layers are created using the models developed, they are 3
4
overlaid together to produce the final flood hazard map of 5
Shind province. The weighted layers were created in order to 6

reduce the number of inputs in the final weighted overlay. 7

Most Suitable
The final output of a flood potential index was the result of
equally weighting 20% each the weighted layers. The results
were validated with the recent flood map of Sindh province. Figure 5. Flood shelter suitability map
Providing flood shelter in flood-affected areas is very
Table 1. District wise flood hazard area
important. Flood shelters provide flood-affected people with
areas to run to for shelter and keeps them safe wh ile the flood Very Very
District High Moderate Low
rages Digital elevation model, land cover, settlement, road high low
and flood map were used to model suitable locations for Larkana 1335 3087 1706 945 439
flood shelter. Preparing the shelters’ suitability slope from Khairpur 1079 1218 4182 9316 0
Jacobabad 944 2788 1840 41 0
the elevation dataset, point density fro m the settlement
Sukkur 941 905 1380 1976 0
dataset, Euclidean distance from the road and flood dataset
Naushahro
were used. Each datasets data were then reclassified to rank 792 1572 666 6 0
Feroze
the suitability of locations for flood shelter. Each raster was Shikarpur 710 1064 770 20 0
assigned a rank at the time of reclassification, with rank ’1’ Ghotki 355 1647 2351 2070 0
as less suitable for flood shelter and ‘5’ as most suitable.
Once all four raster layers have been created using the The 2010 flood in Sindh province, Pakistan was described
models developed, they were overlaid together to produce as the worst environmental disaster in Pakistan's history. The
the final suitability flood shelters location map of Shind flood inundated a total area of 7579 km2 in Sindh province
province (Figure 3). They were then weighted according to a (838 km2 , in Naushahro Feroze district, 965 km2 in Sh ikarpur
percentage of influence and combined to produce a map district, 788 788 km2 in Sukkur d istric, 667 m2 ,in Khairpur
displaying suitable locations for the proposed flood shelter. district, 306 km2 in Ghotki district, 3002 km2 in Jacobabad
 American Journal of Geographic Information System 2013, 2(1): 1-5
district, and 1013 km2 in Larkana district).
The modeled result indicated a very h igh hazard area
(6216 km2 ) out of a total area of 46138 km2 of Sindh
province. The results were validated using the Pakistan flood
2010 data. So me 1363 km2 area was observed to be less flood
hazard-prone and this can be attributed to human
interventions, particularly infrastructure to regularly reroute
water build-up fro m floods.
4. Conclusions
Based on the results of this case study for Sindh province,
it can be concluded that geospatial technology provides the
best potential to analyze and provide results required for
prompt and effective decision-making on floods. This
comprehensive flood hazard and flood shelters model
developed in ArcGIS ModelBuilder can be easily handled by
novice GIS users. The object-based remote sensing
technique and geographical information systems were used
to classify the extent of floods. Object-based classification
systems allow d ifferent rules for d ifferent classes that can be
used further for similar types of flood mapping. The studies
support the conclusion that information derived fro m
different imagery can be very valuable to operational users
for planning flood-related emergency response. Natural
flood disaster is common and cannot be stop. Remote
sensing is an efficient tool for flood mapping and suitability
analysis and can be useful for emergency response and
disaster preparedness.
ACKNOWLEDGEMENTS
We acknowledged the SERVIR-Himalaya in itiative with
the support from USAID and NASA. Our gratitude goes
especially to Mr. Birendra Bajracharya Sr. GIS Specialist,
MENRIS for the encouragement and support extended to
bring this paper.
REFERENCES
[1] “N.M .S.I. Arambepola, G. Iglesias “Effective strategies for
urban flood risk management. Paper presented at the
5
Economic and Social Commission for Asia and the Pacific
Expert Group M eeting on Innovative Strategies towards
Flood Resilient Cities in Asia-Pacific” Bangkok, 21-23 July
2009
[2] International Federation of Red Cross and Red Crescent
Societies. World Disaster Report 1998. Oxford University
Press, 1998
[3] D. Dutta, S. Herath, K. M usiake, “An application of a flood
risk analysis system for impact analysis of a flood control
plan in a river basin Hydrological Processes” Hydrol. Process.
20, 1365–1384, 2006
[4] D. L. Civco, J. D. Hurd, E. H. Wilson, M . Song, Z. Zhang,, “A
comparison of land use and land cover change detection
methods” Proceedings, 2002 A SPRS-ACSM Annual
Conference and FIG XXII Congress,Washington, DC, April
22–26. Bethesda, M D: American Society for
Photogrammetry & Remote Sensing, 22–26, 2002
[5] J.G. Lyon, W.S. Hutchinson, “Application of a radiometric
model for evaluation of water depths and verification of
results with airborne scanner data” Photogrammetric
Engineering and Remote Sensing, 61, 161–166, 1995
[6] W.A. M arcus, C.J. Legleiter, R.J. Aspinall, J.W. Boardman,
R.L. Crabtree, (2003). High spatial resolution hyperspectral
mapping of in-stream habitats, depths, and woody debris in
mountain streams. Geomorphology, 55, 363–380, 2003
[7] R.L. Ketting, D.A. Landgrebe. “Classification of
multispectral image data by extraction and classification of
homogeneous objects” IEEE Transactions on Geoscience
Electronics, GE-14(1): 19–26, 1976
[8] A. Lobo “Image segmentation and discriminant analysis for
the identification of land cover units in ecology” IEEE
Transactions on Geoscience and Remote Sensing, 35(5):
1136–1145, 1997
[9] M . S. Chubey, S. E. Franklin, M . A. Wulder, “Object-based
Analysis of Ikonos-2 Imagery for Extraction of Forest
Inventory Parameters” PhEngRS(72), No. 4 383-394, 2006
[10] J. Harken, R. Sugumaran, “Classification of Iowa wetlands
using an airborne hyperspectral image: A comparison of
spectral angle mapper classifier and an object-oriented
approach”. Canadian Journal of Remote Sensing 31(2):
167–174, 2005
[11] Y. Gao, J. F. M as, I. Niemeyer, P. R. M arpu, J. L. Palacio,
“Object based image analysis for forest area land cover
mapping” In Proceedings, International Symposium for
Spatial Data Quality (ISSDQ). Enschede, the Netherlands,
13–15 June, 2007